API tools faq
paste
Advertisement
Guest User

Untitled

a guest
Jul 18th, 2017
465
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
Diff 48.25 KB | None | 0 0
raw download clone embed print report
  1. From 396a8f247bcf548189de29a7bf9cbad968a7ce50 Mon Sep 17 00:00:00 2001
  2. From: Simon Horlick <simonhorlick@gmail.com>
  3. Date: Tue, 11 Jan 2011 20:05:54 +0000
  4. Subject: [PATCH 1/5] Save interlace decision for all mbs
  5.  
  6. ---
  7. common/common.h     |    1 +
  8.  common/frame.c      |    3 +++
  9.  common/frame.h      |    1 +
  10.  common/macroblock.c |    1 +
  11.  encoder/encoder.c   |    8 ++++++++
  12.  5 files changed, 14 insertions(+), 0 deletions(-)
  13.  
  14. diff --git a/common/common.h b/common/common.h
  15. index 4654c17..b450ac0 100644
  16. --- a/common/common.h
  17. +++ b/common/common.h
  18. @@ -616,6 +616,7 @@ struct x264_t
  19.          int8_t  *mb_transform_size;         /* transform_size_8x8_flag of each mb */
  20.          uint16_t *slice_table;              /* sh->first_mb of the slice that the indexed mb is part of
  21.                                               * NOTE: this will fail on resolutions above 2^16 MBs... */
  22. +        uint8_t *field;
  23.  
  24.           /* buffer for weighted versions of the reference frames */
  25.          pixel *p_weight_buf[X264_REF_MAX];
  26. diff --git a/common/frame.c b/common/frame.c
  27. index ca90539..eff8ca5 100644
  28. --- a/common/frame.c
  29. +++ b/common/frame.c
  30. @@ -145,6 +145,8 @@ x264_frame_t *x264_frame_new( x264_t *h, int b_fdec )
  31.                              frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv) * sizeof(uint16_t) << h->frames.b_have_sub8x8_esa );
  32.              frame->integral = (uint16_t*)frame->buffer[3] + frame->i_stride[0] * i_padv + PADH;
  33.          }
  34. +        if( h->param.b_interlaced )
  35. +            CHECKED_MALLOC( frame->field, i_mb_count * sizeof(uint8_t) );
  36.      }
  37.      else /* fenc frame */
  38.      {
  39. @@ -219,6 +221,7 @@ void x264_frame_delete( x264_frame_t *frame )
  40.          x264_free( frame->i_inv_qscale_factor );
  41.          x264_free( frame->i_row_bits );
  42.          x264_free( frame->f_row_qp );
  43. +        x264_free( frame->field );
  44.          x264_free( frame->mb_type );
  45.          x264_free( frame->mb_partition );
  46.          x264_free( frame->mv[0] );
  47. diff --git a/common/frame.h b/common/frame.h
  48. index 38d0bf2..7a1bd77 100644
  49. --- a/common/frame.h
  50. +++ b/common/frame.h
  51. @@ -62,6 +62,7 @@ typedef struct x264_frame
  52.      float   f_qp_avg_rc; /* QPs as decided by ratecontrol */
  53.      float   f_qp_avg_aq; /* QPs as decided by AQ in addition to ratecontrol */
  54.      int     i_poc_l0ref0; /* poc of first refframe in L0, used to check if direct temporal is possible */
  55. +    uint8_t *field;
  56.  
  57.      /* YUV buffer */
  58.      int     i_plane;
  59. diff --git a/common/macroblock.c b/common/macroblock.c
  60. index e559ab1..de5b01f 100644
  61. --- a/common/macroblock.c
  62. +++ b/common/macroblock.c
  63. @@ -368,6 +368,7 @@ void x264_macroblock_slice_init( x264_t *h )
  64.      h->mb.ref[1] = h->fdec->ref[1];
  65.      h->mb.type = h->fdec->mb_type;
  66.      h->mb.partition = h->fdec->mb_partition;
  67. +    h->mb.field = h->fdec->field;
  68.  
  69.      h->fdec->i_ref[0] = h->i_ref[0];
  70.      h->fdec->i_ref[1] = h->i_ref[1];
  71. diff --git a/encoder/encoder.c b/encoder/encoder.c
  72. index 0182c8c..3b9a175 100644
  73. --- a/encoder/encoder.c
  74. +++ b/encoder/encoder.c
  75. @@ -1950,6 +1950,14 @@ static int x264_slice_write( x264_t *h )
  76.              }
  77.          }
  78.  
  79. +        if( h->param.b_interlaced )
  80. +        {
  81. +            if( !(i_mb_y&1) )
  82. +                h->mb.b_interlaced = 1;
  83. +            x264_zigzag_init( h->param.cpu, &h->zigzagf, h->mb.b_interlaced );
  84. +            h->mb.field[mb_xy] = h->mb.b_interlaced;
  85. +        }
  86. +
  87.          if( i_mb_x == 0 && !h->mb.b_reencode_mb )
  88.              x264_fdec_filter_row( h, i_mb_y, 1 );
  89.  
  90. --
  91. 1.7.1
  92.  
  93.  
  94. From 05caa9ab13813eb34aca6e01841ef975576db4bc Mon Sep 17 00:00:00 2001
  95. From: Simon Horlick <simonhorlick@gmail.com>
  96. Date: Tue, 11 Jan 2011 20:09:00 +0000
  97. Subject: [PATCH 2/5] Disable adaptive MBAFF when subme 0 is used.
  98.  
  99. ---
  100. encoder/encoder.c |    9 ++++++---
  101.  x264.c            |   12 ++++++++++++
  102.  x264.h            |    1 +
  103.  3 files changed, 19 insertions(+), 3 deletions(-)
  104.  
  105. diff --git a/encoder/encoder.c b/encoder/encoder.c
  106. index 3b9a175..4fb0a57 100644
  107. --- a/encoder/encoder.c
  108. +++ b/encoder/encoder.c
  109. @@ -1952,9 +1952,12 @@ static int x264_slice_write( x264_t *h )
  110.  
  111.          if( h->param.b_interlaced )
  112.          {
  113. -            if( !(i_mb_y&1) )
  114. -                h->mb.b_interlaced = 1;
  115. -            x264_zigzag_init( h->param.cpu, &h->zigzagf, h->mb.b_interlaced );
  116. +            if( h->param.b_adaptive_mbaff )
  117. +            {
  118. +                if( !(i_mb_y&1) )
  119. +                    h->mb.b_interlaced = 1;
  120. +                x264_zigzag_init( h->param.cpu, &h->zigzagf, h->mb.b_interlaced );
  121. +            }
  122.              h->mb.field[mb_xy] = h->mb.b_interlaced;
  123.          }
  124.  
  125. diff --git a/x264.c b/x264.c
  126. index c6261d8..d9c60cd 100644
  127. --- a/x264.c
  128. +++ b/x264.c
  129. @@ -1494,6 +1494,18 @@ generic_option:
  130.          param->b_tff = !!info.tff;
  131.      }
  132.  
  133. +    /* Adaptive MBAFF and subme 0 are not supported as motion vectors between
  134. +     * field macroblocks and frame macroblocks require halving and hpel pixels.
  135. +     * The chosen solution is to make MBAFF non-adaptive in this case. */
  136. +    if( param->b_interlaced && !param->analyse.i_subpel_refine )
  137. +    {
  138. +        x264_cli_log( "x264", X264_LOG_WARNING, "Adaptive MBAFF and subme 0 are not supported. "
  139. +                "Using non-adaptive MBAFF instead.\n");
  140. +        param->b_adaptive_mbaff = 0;
  141. +    }
  142. +    else
  143. +        param->b_adaptive_mbaff = 1;
  144. +
  145.      /* Automatically reduce reference frame count to match the user's target level
  146.       * if the user didn't explicitly set a reference frame count. */
  147.      if( !b_user_ref )
  148. diff --git a/x264.h b/x264.h
  149. index 8f39497..739d456 100644
  150. --- a/x264.h
  151. +++ b/x264.h
  152. @@ -291,6 +291,7 @@ typedef struct x264_param_t
  153.      int         i_cabac_init_idc;
  154.  
  155.      int         b_interlaced;
  156. +    int         b_adaptive_mbaff; /* MBAFF+subme 0 require non-adaptive MBAFF i.e. all field mbs */
  157.      int         b_constrained_intra;
  158.  
  159.      int         i_cqm_preset;
  160. --
  161. 1.7.1
  162.  
  163.  
  164. From ac56eec487176b63ca67c20da5e04d96faba9652 Mon Sep 17 00:00:00 2001
  165. From: Simon Horlick <simonhorlick@gmail.com>
  166. Date: Tue, 11 Jan 2011 20:16:18 +0000
  167. Subject: [PATCH 3/5] Left index table
  168.  
  169. ---
  170. common/common.h     |    1 +
  171.  common/macroblock.c |   56 ++++++++++++++++++++++++++++++++------------------
  172.  2 files changed, 37 insertions(+), 20 deletions(-)
  173.  
  174. diff --git a/common/common.h b/common/common.h
  175. index b450ac0..f244e3d 100644
  176. --- a/common/common.h
  177. +++ b/common/common.h
  178. @@ -594,6 +594,7 @@ struct x264_t
  179.          int     i_mb_top_xy;
  180.          int     i_mb_topleft_xy;
  181.          int     i_mb_topright_xy;
  182. +        int     *left_index_table;
  183.  
  184.          /**** thread synchronization ends here ****/
  185.          /* subsequent variables are either thread-local or constant,
  186. diff --git a/common/macroblock.c b/common/macroblock.c
  187. index de5b01f..40f5c67 100644
  188. --- a/common/macroblock.c
  189. +++ b/common/macroblock.c
  190. @@ -547,6 +547,18 @@ static void ALWAYS_INLINE x264_macroblock_load_pic_pointers( x264_t *h, int mb_x
  191.          }
  192.  }
  193.  
  194. +static const int left_indices[5][22] = {
  195. +/*    intra modes     nnz                                          mv          ref    real indices */
  196. +    /* Current is progressive. */
  197. +    { 4, 4, 5, 5,     3,  3,  7,  7, 16+1, 16+1, 16+4+1, 16+4+1,   0, 0, 1, 1, 0, 0,  0, 0, 1, 1 },
  198. +    { 6, 6, 3, 3,    11, 11, 15, 15, 16+3, 16+3, 16+4+3, 16+4+3,   2, 2, 3, 3, 1, 1,  2, 2, 3, 3 },
  199. +    /* Current is interlaced.*/
  200. +    { 4, 6, 4, 6,     3, 11,  3, 11, 16+1, 16+1, 16+4+1, 16+4+1,   0, 2, 0, 2, 0, 0,  0, 2, 0, 2 },
  201. +    { 4, 6, 4, 6,     3, 11,  3, 11, 16+1, 16+1, 16+4+1, 16+4+1,   0, 2, 0, 2, 0, 0,  0, 2, 0, 2 },
  202. +    /*Both same.*/
  203. +    { 4, 5, 6, 3,     3,  7, 11, 15, 16+1, 16+3, 16+4+1, 16+4+3,   0, 1, 2, 3, 0, 1,  0, 1, 2, 3 },
  204. +};
  205. +
  206.  static void inline x264_macroblock_cache_load_neighbours( x264_t *h, int mb_x, int mb_y )
  207.  {
  208.      int top = (mb_y - (1 << h->mb.b_interlaced)) * h->mb.i_mb_stride + mb_x;
  209. @@ -567,6 +579,7 @@ static void inline x264_macroblock_cache_load_neighbours( x264_t *h, int mb_x, i
  210.      h->mb.i_mb_type_left = -1;
  211.      h->mb.i_mb_type_topleft = -1;
  212.      h->mb.i_mb_type_topright = -1;
  213. +    h->mb.left_index_table = left_indices[4];
  214.  
  215.      if( mb_x > 0 )
  216.      {
  217. @@ -658,6 +671,8 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  218.      uint8_t (*nnz)[24] = h->mb.non_zero_count;
  219.      int16_t *cbp = h->mb.cbp;
  220.  
  221. +    int *left_index_table = h->mb.left_index_table;
  222. +
  223.      /* load cache */
  224.      if( h->mb.i_neighbour & MB_TOP )
  225.      {
  226. @@ -700,22 +715,22 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  227.          h->mb.cache.i_cbp_left = cbp[left];
  228.  
  229.          /* load intra4x4 */
  230. -        h->mb.cache.intra4x4_pred_mode[x264_scan8[0 ] - 1] = i4x4[left][4];
  231. -        h->mb.cache.intra4x4_pred_mode[x264_scan8[2 ] - 1] = i4x4[left][5];
  232. -        h->mb.cache.intra4x4_pred_mode[x264_scan8[8 ] - 1] = i4x4[left][6];
  233. -        h->mb.cache.intra4x4_pred_mode[x264_scan8[10] - 1] = i4x4[left][3];
  234. +        h->mb.cache.intra4x4_pred_mode[x264_scan8[0 ] - 1] = i4x4[left][left_index_table[0]];
  235. +        h->mb.cache.intra4x4_pred_mode[x264_scan8[2 ] - 1] = i4x4[left][left_index_table[1]];
  236. +        h->mb.cache.intra4x4_pred_mode[x264_scan8[8 ] - 1] = i4x4[left][left_index_table[2]];
  237. +        h->mb.cache.intra4x4_pred_mode[x264_scan8[10] - 1] = i4x4[left][left_index_table[3]];
  238.  
  239.          /* load non_zero_count */
  240. -        h->mb.cache.non_zero_count[x264_scan8[0 ] - 1] = nnz[left][3];
  241. -        h->mb.cache.non_zero_count[x264_scan8[2 ] - 1] = nnz[left][7];
  242. -        h->mb.cache.non_zero_count[x264_scan8[8 ] - 1] = nnz[left][11];
  243. -        h->mb.cache.non_zero_count[x264_scan8[10] - 1] = nnz[left][15];
  244. +        h->mb.cache.non_zero_count[x264_scan8[0 ] - 1] = nnz[left][left_index_table[4+0]];
  245. +        h->mb.cache.non_zero_count[x264_scan8[2 ] - 1] = nnz[left][left_index_table[4+1]];
  246. +        h->mb.cache.non_zero_count[x264_scan8[8 ] - 1] = nnz[left][left_index_table[4+2]];
  247. +        h->mb.cache.non_zero_count[x264_scan8[10] - 1] = nnz[left][left_index_table[4+3]];
  248.  
  249. -        h->mb.cache.non_zero_count[x264_scan8[16+0] - 1] = nnz[left][16+1];
  250. -        h->mb.cache.non_zero_count[x264_scan8[16+2] - 1] = nnz[left][16+3];
  251. +        h->mb.cache.non_zero_count[x264_scan8[16+0] - 1] = nnz[left][left_index_table[4+4]];
  252. +        h->mb.cache.non_zero_count[x264_scan8[16+2] - 1] = nnz[left][left_index_table[4+5]];
  253.  
  254. -        h->mb.cache.non_zero_count[x264_scan8[16+4+0] - 1] = nnz[left][16+4+1];
  255. -        h->mb.cache.non_zero_count[x264_scan8[16+4+2] - 1] = nnz[left][16+4+3];
  256. +        h->mb.cache.non_zero_count[x264_scan8[16+4+0] - 1] = nnz[left][left_index_table[4+6]];
  257. +        h->mb.cache.non_zero_count[x264_scan8[16+4+2] - 1] = nnz[left][left_index_table[4+7]];
  258.      }
  259.      else
  260.      {
  261. @@ -854,10 +869,10 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  262.  
  263.              if( h->mb.i_neighbour & MB_LEFT )
  264.              {
  265. -                CP16( h->mb.cache.mvd[l][x264_scan8[0 ] - 1], mvd[left][4] );
  266. -                CP16( h->mb.cache.mvd[l][x264_scan8[2 ] - 1], mvd[left][5] );
  267. -                CP16( h->mb.cache.mvd[l][x264_scan8[8 ] - 1], mvd[left][6] );
  268. -                CP16( h->mb.cache.mvd[l][x264_scan8[10] - 1], mvd[left][3] );
  269. +                CP16( h->mb.cache.mvd[l][x264_scan8[0 ] - 1], mvd[left][left_index_table[0]] );
  270. +                CP16( h->mb.cache.mvd[l][x264_scan8[2 ] - 1], mvd[left][left_index_table[1]] );
  271. +                CP16( h->mb.cache.mvd[l][x264_scan8[8 ] - 1], mvd[left][left_index_table[2]] );
  272. +                CP16( h->mb.cache.mvd[l][x264_scan8[10] - 1], mvd[left][left_index_table[3]] );
  273.              }
  274.              else
  275.                  for( int i = 0; i < 4; i++ )
  276. @@ -946,6 +961,7 @@ void x264_macroblock_cache_load_deblock( x264_t *h )
  277.              int s4x4 = h->mb.i_b4_stride;
  278.  
  279.              uint8_t (*nnz)[24] = h->mb.non_zero_count;
  280. +            int *left_index_table = h->mb.left_index_table;
  281.  
  282.              if( h->mb.i_neighbour & MB_TOP )
  283.                  CP32( &h->mb.cache.non_zero_count[x264_scan8[0] - 8], &nnz[h->mb.i_mb_top_xy][12] );
  284. @@ -953,10 +969,10 @@ void x264_macroblock_cache_load_deblock( x264_t *h )
  285.              if( h->mb.i_neighbour & MB_LEFT )
  286.              {
  287.                  int left = h->mb.i_mb_left_xy;
  288. -                h->mb.cache.non_zero_count[x264_scan8[0 ] - 1] = nnz[left][3];
  289. -                h->mb.cache.non_zero_count[x264_scan8[2 ] - 1] = nnz[left][7];
  290. -                h->mb.cache.non_zero_count[x264_scan8[8 ] - 1] = nnz[left][11];
  291. -                h->mb.cache.non_zero_count[x264_scan8[10] - 1] = nnz[left][15];
  292. +                h->mb.cache.non_zero_count[x264_scan8[0 ] - 1] = nnz[left][left_index_table[4+0]];
  293. +                h->mb.cache.non_zero_count[x264_scan8[2 ] - 1] = nnz[left][left_index_table[4+1]];
  294. +                h->mb.cache.non_zero_count[x264_scan8[8 ] - 1] = nnz[left][left_index_table[4+2]];
  295. +                h->mb.cache.non_zero_count[x264_scan8[10] - 1] = nnz[left][left_index_table[4+3]];
  296.              }
  297.  
  298.              for( int l = 0; l <= (h->sh.i_type == SLICE_TYPE_B); l++ )
  299. --
  300. 1.7.1
  301.  
  302.  
  303. From 8e33318d3a7c64c2f4a8e77bed4464e5748caf33 Mon Sep 17 00:00:00 2001
  304. From: Simon Horlick <simonhorlick@gmail.com>
  305. Date: Tue, 11 Jan 2011 20:21:26 +0000
  306. Subject: [PATCH 4/5] Store references to the two left macroblocks in MBAFF
  307.  
  308. ---
  309. common/common.h     |    4 ++--
  310.  common/deblock.c    |    4 ++--
  311.  common/macroblock.c |   20 ++++++++++----------
  312.  common/mvpred.c     |    2 +-
  313.  encoder/analyse.c   |    8 ++++----
  314.  encoder/cabac.c     |    8 ++++----
  315.  6 files changed, 23 insertions(+), 23 deletions(-)
  316.  
  317. diff --git a/common/common.h b/common/common.h
  318. index f244e3d..d49c41f 100644
  319. --- a/common/common.h
  320. +++ b/common/common.h
  321. @@ -586,11 +586,11 @@ struct x264_t
  322.          unsigned int i_neighbour_intra;     /* for constrained intra pred */
  323.          unsigned int i_neighbour_frame;     /* ignoring slice boundaries */
  324.          int     i_mb_type_top;
  325. -        int     i_mb_type_left;
  326. +        int     i_mb_type_left[2];
  327.          int     i_mb_type_topleft;
  328.          int     i_mb_type_topright;
  329.          int     i_mb_prev_xy;
  330. -        int     i_mb_left_xy;
  331. +        int     i_mb_left_xy[2];
  332.          int     i_mb_top_xy;
  333.          int     i_mb_topleft_xy;
  334.          int     i_mb_topright_xy;
  335. diff --git a/common/deblock.c b/common/deblock.c
  336. index ff7c99f..5074b77 100644
  337. --- a/common/deblock.c
  338. +++ b/common/deblock.c
  339. @@ -347,10 +347,10 @@ void x264_frame_deblock_row( x264_t *h, int mb_y )
  340.  
  341.          if( h->mb.i_neighbour & MB_LEFT )
  342.          {
  343. -            int qpl = h->mb.qp[h->mb.i_mb_left_xy];
  344. +            int qpl = h->mb.qp[h->mb.i_mb_left_xy[0]];
  345.              int qp_left = (qp + qpl + 1) >> 1;
  346.              int qpc_left = (h->chroma_qp_table[qp] + h->chroma_qp_table[qpl] + 1) >> 1;
  347. -            int intra_left = IS_INTRA( h->mb.type[h->mb.i_mb_left_xy] );
  348. +            int intra_left = IS_INTRA( h->mb.type[h->mb.i_mb_left_xy[0]] );
  349.              if( intra_cur || intra_left )
  350.                  FILTER( _intra, 0, 0, qp_left, qpc_left );
  351.              else
  352. diff --git a/common/macroblock.c b/common/macroblock.c
  353. index 40f5c67..1e40b28 100644
  354. --- a/common/macroblock.c
  355. +++ b/common/macroblock.c
  356. @@ -572,11 +572,11 @@ static void inline x264_macroblock_cache_load_neighbours( x264_t *h, int mb_x, i
  357.      h->mb.i_neighbour_intra = 0;
  358.      h->mb.i_neighbour_frame = 0;
  359.      h->mb.i_mb_top_xy = -1;
  360. -    h->mb.i_mb_left_xy = -1;
  361. +    h->mb.i_mb_left_xy[0] = h->mb.i_mb_left_xy[1] = -1;
  362.      h->mb.i_mb_topleft_xy = -1;
  363.      h->mb.i_mb_topright_xy = -1;
  364.      h->mb.i_mb_type_top = -1;
  365. -    h->mb.i_mb_type_left = -1;
  366. +    h->mb.i_mb_type_left[0] = h->mb.i_mb_type_left[1] = -1;
  367.      h->mb.i_mb_type_topleft = -1;
  368.      h->mb.i_mb_type_topright = -1;
  369.      h->mb.left_index_table = left_indices[4];
  370. @@ -584,13 +584,13 @@ static void inline x264_macroblock_cache_load_neighbours( x264_t *h, int mb_x, i
  371.      if( mb_x > 0 )
  372.      {
  373.          h->mb.i_neighbour_frame |= MB_LEFT;
  374. -        h->mb.i_mb_left_xy = h->mb.i_mb_xy - 1;
  375. -        h->mb.i_mb_type_left = h->mb.type[h->mb.i_mb_left_xy];
  376. +        h->mb.i_mb_left_xy[0] = h->mb.i_mb_xy - 1;
  377. +        h->mb.i_mb_type_left[0] = h->mb.type[h->mb.i_mb_left_xy[0]];
  378.          if( h->mb.i_mb_xy > h->sh.i_first_mb )
  379.          {
  380.              h->mb.i_neighbour |= MB_LEFT;
  381.  
  382. -            if( !h->param.b_constrained_intra || IS_INTRA( h->mb.i_mb_type_left ) )
  383. +            if( !h->param.b_constrained_intra || IS_INTRA( h->mb.i_mb_type_left[0] ) )
  384.                  h->mb.i_neighbour_intra |= MB_LEFT;
  385.          }
  386.      }
  387. @@ -656,7 +656,7 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  388.  {
  389.      x264_macroblock_cache_load_neighbours( h, mb_x, mb_y );
  390.  
  391. -    int left = h->mb.i_mb_left_xy;
  392. +    int left = h->mb.i_mb_left_xy[0];
  393.      int top  = h->mb.i_mb_top_xy;
  394.      int top_y = mb_y - (1 << h->mb.b_interlaced);
  395.      int s8x8 = h->mb.i_b8_stride;
  396. @@ -924,8 +924,8 @@ void x264_macroblock_cache_load_neighbours_deblock( x264_t *h, int mb_x, int mb_
  397.  
  398.      if( mb_x > 0 )
  399.      {
  400. -        h->mb.i_mb_left_xy = h->mb.i_mb_xy - 1;
  401. -        if( deblock_on_slice_edges || h->mb.slice_table[h->mb.i_mb_left_xy] == h->mb.slice_table[h->mb.i_mb_xy] )
  402. +        h->mb.i_mb_left_xy[0] = h->mb.i_mb_xy - 1;
  403. +        if( deblock_on_slice_edges || h->mb.slice_table[h->mb.i_mb_left_xy[0]] == h->mb.slice_table[h->mb.i_mb_xy] )
  404.              h->mb.i_neighbour |= MB_LEFT;
  405.      }
  406.  
  407. @@ -968,7 +968,7 @@ void x264_macroblock_cache_load_deblock( x264_t *h )
  408.  
  409.              if( h->mb.i_neighbour & MB_LEFT )
  410.              {
  411. -                int left = h->mb.i_mb_left_xy;
  412. +                int left = h->mb.i_mb_left_xy[0];
  413.                  h->mb.cache.non_zero_count[x264_scan8[0 ] - 1] = nnz[left][left_index_table[4+0]];
  414.                  h->mb.cache.non_zero_count[x264_scan8[2 ] - 1] = nnz[left][left_index_table[4+1]];
  415.                  h->mb.cache.non_zero_count[x264_scan8[8 ] - 1] = nnz[left][left_index_table[4+2]];
  416. @@ -1043,7 +1043,7 @@ void x264_macroblock_cache_load_deblock( x264_t *h )
  417.      {
  418.          uint8_t (*nnz)[24] = h->mb.non_zero_count;
  419.          int top = h->mb.i_mb_top_xy;
  420. -        int left = h->mb.i_mb_left_xy;
  421. +        int left = h->mb.i_mb_left_xy[0];
  422.  
  423.          if( (h->mb.i_neighbour & MB_TOP) && h->mb.mb_transform_size[top] )
  424.          {
  425. diff --git a/common/mvpred.c b/common/mvpred.c
  426. index a24dde8..c8efe1f 100644
  427. --- a/common/mvpred.c
  428. +++ b/common/mvpred.c
  429. @@ -426,7 +426,7 @@ void x264_mb_predict_mv_ref16x16( x264_t *h, int i_list, int i_ref, int16_t mvc[
  430.      }
  431.  
  432.      /* spatial predictors */
  433. -    SET_MVP( mvr[h->mb.i_mb_left_xy] );
  434. +    SET_MVP( mvr[h->mb.i_mb_left_xy[0]] );
  435.      SET_MVP( mvr[h->mb.i_mb_top_xy] );
  436.      SET_MVP( mvr[h->mb.i_mb_topleft_xy] );
  437.      SET_MVP( mvr[h->mb.i_mb_topright_xy] );
  438. diff --git a/encoder/analyse.c b/encoder/analyse.c
  439. index bf9f3c9..d87950e 100644
  440. --- a/encoder/analyse.c
  441. +++ b/encoder/analyse.c
  442. @@ -516,7 +516,7 @@ static void x264_mb_analyse_init( x264_t *h, x264_mb_analysis_t *a, int qp )
  443.          {
  444.              /* Always run in fast-intra mode for subme < 3 */
  445.              if( h->mb.i_subpel_refine > 2 &&
  446. -              ( IS_INTRA( h->mb.i_mb_type_left ) ||
  447. +              ( IS_INTRA( h->mb.i_mb_type_left[0] ) ||
  448.                  IS_INTRA( h->mb.i_mb_type_top ) ||
  449.                  IS_INTRA( h->mb.i_mb_type_topleft ) ||
  450.                  IS_INTRA( h->mb.i_mb_type_topright ) ||
  451. @@ -1296,7 +1296,7 @@ static void x264_mb_analyse_inter_p8x8_mixed_ref( x264_t *h, x264_mb_analysis_t
  452.      /* early termination: if 16x16 chose ref 0, then evalute no refs older
  453.       * than those used by the neighbors */
  454.      if( i_maxref > 0 && (a->l0.me16x16.i_ref == 0 || a->l0.me16x16.i_ref == h->mb.ref_blind_dupe) &&
  455. -        h->mb.i_mb_type_top > 0 && h->mb.i_mb_type_left > 0 )
  456. +        h->mb.i_mb_type_top > 0 && h->mb.i_mb_type_left[0] > 0 )
  457.      {
  458.          i_maxref = 0;
  459.          CHECK_NEIGHBOUR(  -8 - 1 );
  460. @@ -2063,7 +2063,7 @@ static void x264_mb_analyse_inter_b8x8_mixed_ref( x264_t *h, x264_mb_analysis_t
  461.      {
  462.          x264_mb_analysis_list_t *lX = l ? &a->l1 : &a->l0;
  463.          if( i_maxref[l] > 0 && lX->me16x16.i_ref == 0 &&
  464. -            h->mb.i_mb_type_top > 0 && h->mb.i_mb_type_left > 0 )
  465. +            h->mb.i_mb_type_top > 0 && h->mb.i_mb_type_left[0] > 0 )
  466.          {
  467.              i_maxref[l] = 0;
  468.              CHECK_NEIGHBOUR(  -8 - 1 );
  469. @@ -2817,7 +2817,7 @@ intra_analysis:
  470.                      {}
  471.                  else if( h->param.analyse.i_subpel_refine >= 3 )
  472.                      analysis.b_try_skip = 1;
  473. -                else if( h->mb.i_mb_type_left == P_SKIP ||
  474. +                else if( h->mb.i_mb_type_left[0] == P_SKIP ||
  475.                           h->mb.i_mb_type_top == P_SKIP ||
  476.                           h->mb.i_mb_type_topleft == P_SKIP ||
  477.                           h->mb.i_mb_type_topright == P_SKIP )
  478. diff --git a/encoder/cabac.c b/encoder/cabac.c
  479. index 6333737..334318d 100644
  480. --- a/encoder/cabac.c
  481. +++ b/encoder/cabac.c
  482. @@ -79,7 +79,7 @@ static void x264_cabac_mb_type( x264_t *h, x264_cabac_t *cb )
  483.      if( h->sh.i_type == SLICE_TYPE_I )
  484.      {
  485.          int ctx = 0;
  486. -        if( (h->mb.i_neighbour & MB_LEFT) && h->mb.i_mb_type_left != I_4x4 )
  487. +        if( (h->mb.i_neighbour & MB_LEFT) && h->mb.i_mb_type_left[0] != I_4x4 )
  488.              ctx++;
  489.          if( (h->mb.i_neighbour & MB_TOP) && h->mb.i_mb_type_top != I_4x4 )
  490.              ctx++;
  491. @@ -113,7 +113,7 @@ static void x264_cabac_mb_type( x264_t *h, x264_cabac_t *cb )
  492.      else //if( h->sh.i_type == SLICE_TYPE_B )
  493.      {
  494.          int ctx = 0;
  495. -        if( (h->mb.i_neighbour & MB_LEFT) && h->mb.i_mb_type_left != B_SKIP && h->mb.i_mb_type_left != B_DIRECT )
  496. +        if( (h->mb.i_neighbour & MB_LEFT) && h->mb.i_mb_type_left[0] != B_SKIP && h->mb.i_mb_type_left[0] != B_DIRECT )
  497.              ctx++;
  498.          if( (h->mb.i_neighbour & MB_TOP) && h->mb.i_mb_type_top != B_SKIP && h->mb.i_mb_type_top != B_DIRECT )
  499.              ctx++;
  500. @@ -198,7 +198,7 @@ static void x264_cabac_mb_intra_chroma_pred_mode( x264_t *h, x264_cabac_t *cb )
  501.      int       ctx = 0;
  502.  
  503.      /* No need to test for I4x4 or I_16x16 as cache_save handle that */
  504. -    if( (h->mb.i_neighbour & MB_LEFT) && h->mb.chroma_pred_mode[h->mb.i_mb_left_xy] != 0 )
  505. +    if( (h->mb.i_neighbour & MB_LEFT) && h->mb.chroma_pred_mode[h->mb.i_mb_left_xy[0]] != 0 )
  506.          ctx++;
  507.      if( (h->mb.i_neighbour & MB_TOP) && h->mb.chroma_pred_mode[h->mb.i_mb_top_xy] != 0 )
  508.          ctx++;
  509. @@ -280,7 +280,7 @@ static void x264_cabac_mb_qp_delta( x264_t *h, x264_cabac_t *cb )
  510.  #if !RDO_SKIP_BS
  511.  void x264_cabac_mb_skip( x264_t *h, int b_skip )
  512.  {
  513. -    int ctx = ((h->mb.i_neighbour & MB_LEFT) && !IS_SKIP( h->mb.i_mb_type_left ))
  514. +    int ctx = ((h->mb.i_neighbour & MB_LEFT) && !IS_SKIP( h->mb.i_mb_type_left[0] ))
  515.              + ((h->mb.i_neighbour & MB_TOP) && !IS_SKIP( h->mb.i_mb_type_top ))
  516.              + (h->sh.i_type == SLICE_TYPE_P ? 11 : 24);
  517.      x264_cabac_encode_decision( &h->cabac, ctx, b_skip );
  518. --
  519. 1.7.1
  520.  
  521.  
  522. From 988b2632ea8b8a98f385840313284425d8f0cd89 Mon Sep 17 00:00:00 2001
  523. From: Simon Horlick <simonhorlick@gmail.com>
  524. Date: Fri, 14 Jan 2011 21:18:14 +0000
  525. Subject: [PATCH 5/5] Back up intra borders correctly for MBAFF and make neighbour calculation several times longer.
  526.  
  527. ---
  528. common/common.h     |    9 ++-
  529.  common/macroblock.c |  292 +++++++++++++++++++++++++++++++++++++++++++--------
  530.  2 files changed, 255 insertions(+), 46 deletions(-)
  531.  
  532. diff --git a/common/common.h b/common/common.h
  533. index d49c41f..913c6f0 100644
  534. --- a/common/common.h
  535. +++ b/common/common.h
  536. @@ -548,6 +548,8 @@ struct x264_t
  537.          int     i_mb_stride;
  538.          int     i_b8_stride;
  539.          int     i_b4_stride;
  540. +        int     left_b8[2];
  541. +        int     left_b4[2];
  542.  
  543.          /* Current index */
  544.          int     i_mb_x;
  545. @@ -595,6 +597,10 @@ struct x264_t
  546.          int     i_mb_topleft_xy;
  547.          int     i_mb_topright_xy;
  548.          int     *left_index_table;
  549. +        int     topleft_partition;
  550. +        int     intra_border_index;
  551. +        int     topleft_border_index;
  552. +        int     topright_border_index;
  553.  
  554.          /**** thread synchronization ends here ****/
  555.          /* subsequent variables are either thread-local or constant,
  556. @@ -825,7 +831,8 @@ struct x264_t
  557.  
  558.      /* Buffers that are allocated per-thread even in sliced threads. */
  559.      void *scratch_buffer; /* for any temporary storage that doesn't want repeated malloc */
  560. -    pixel *intra_border_backup[2][2]; /* bottom pixels of the previous mb row, used for intra prediction after the framebuffer has been deblocked */
  561. +    pixel *intra_border_backup[3][2]; /* bottom pixels of the previous mb row, used for intra prediction after the framebuffer has been deblocked */
  562. +    pixel *intra_diagonal_backup[5][3];
  563.      uint8_t (*deblock_strength[2])[2][4][4];
  564.  
  565.      /* CPU functions dependents */
  566. diff --git a/common/macroblock.c b/common/macroblock.c
  567. index 1e40b28..5c2ffc4 100644
  568. --- a/common/macroblock.c
  569. +++ b/common/macroblock.c
  570. @@ -314,18 +314,35 @@ void x264_macroblock_cache_free( x264_t *h )
  571.  int x264_macroblock_thread_allocate( x264_t *h, int b_lookahead )
  572.  {
  573.      if( !b_lookahead )
  574. -        for( int i = 0; i <= h->param.b_interlaced; i++ )
  575. +    {
  576. +        for( int i = 0; i <= 2*h->param.b_interlaced; i++ )
  577.          {
  578.              for( int j = 0; j < 2; j++ )
  579.              {
  580.                  /* shouldn't really be initialized, just silences a valgrind false-positive in predict_8x8_filter_mmx */
  581.                  CHECKED_MALLOCZERO( h->intra_border_backup[i][j], (h->sps->i_mb_width*16+32) * sizeof(pixel) );
  582.                  h->intra_border_backup[i][j] += 16;
  583. -                h->intra_border_backup[1][j] = h->intra_border_backup[i][j];
  584. +                if( !h->param.b_interlaced )
  585. +                    h->intra_border_backup[1][j] = h->intra_border_backup[i][j];
  586. +            }
  587. +        }
  588. +        for( int i = 0; i < 4*h->mb.b_interlaced; i++ )
  589. +        {
  590. +            for( int j = 0; j < 3; j++ )
  591. +            {
  592. +                const int width = 1 + 8; // top left pixel + eight top right pixels (for luma)
  593. +                CHECKED_MALLOCZERO( h->intra_diagonal_backup[i][j], (h->sps->i_mb_width*width+32) * sizeof(pixel) );
  594. +                h->intra_diagonal_backup[i][j] += 16;
  595. +                if( !h->param.b_interlaced )
  596. +                    h->intra_diagonal_backup[1][j] = h->intra_diagonal_backup[i][j];
  597.              }
  598. +        }
  599. +        for( int i = 0; i <= h->param.b_interlaced; i++ )
  600. +        {
  601.              CHECKED_MALLOC( h->deblock_strength[i], sizeof(**h->deblock_strength) * h->mb.i_mb_width );
  602.              h->deblock_strength[1] = h->deblock_strength[i];
  603.          }
  604. +    }
  605.  
  606.      /* Allocate scratch buffer */
  607.      int scratch_size = 0;
  608. @@ -350,12 +367,20 @@ fail:
  609.  void x264_macroblock_thread_free( x264_t *h, int b_lookahead )
  610.  {
  611.      if( !b_lookahead )
  612. +    {
  613.          for( int i = 0; i <= h->param.b_interlaced; i++ )
  614. -        {
  615.              x264_free( h->deblock_strength[i] );
  616. +        for( int i = 0; i <= 2*h->param.b_interlaced; i++ )
  617. +        {
  618.              for( int j = 0; j < 2; j++ )
  619.                  x264_free( h->intra_border_backup[i][j] - 16 );
  620.          }
  621. +        for( int i = 0; i < 4*h->param.b_interlaced; i++ )
  622. +        {
  623. +            for( int j = 0; j < 3; j++ )
  624. +                x264_free( h->intra_diagonal_backup[i][j] - 16 );
  625. +        }
  626. +    }
  627.      x264_free( h->scratch_buffer );
  628.  }
  629.  
  630. @@ -491,14 +516,15 @@ static void ALWAYS_INLINE x264_macroblock_load_pic_pointers( x264_t *h, int mb_x
  631.  {
  632.      int w = (i ? 8 : 16);
  633.      int i_stride = h->fdec->i_stride[i];
  634. -    int i_stride2 = i_stride << b_interlaced;
  635. -    int i_pix_offset = b_interlaced
  636. +    int i_stride2 = i_stride << h->mb.b_interlaced;
  637. +    int i_pix_offset = h->mb.b_interlaced
  638.                       ? 16 * mb_x + w * (mb_y&~1) * i_stride + (mb_y&1) * i_stride
  639.                       : 16 * mb_x + w * mb_y * i_stride;
  640.      pixel *plane_fdec = &h->fdec->plane[i][i_pix_offset];
  641. -    pixel *intra_fdec = &h->intra_border_backup[mb_y&1][i][mb_x*16];
  642. +    pixel *intra_fdec = &h->intra_border_backup[h->mb.intra_border_index][i][mb_x*16];
  643.      int ref_pix_offset[2] = { i_pix_offset, i_pix_offset };
  644. -    if( b_interlaced )
  645. +    /* ref_pix_offset[0] references the current field and [1] the opposite field. */
  646. +    if( h->mb.b_interlaced )
  647.          ref_pix_offset[1] += (1-2*(mb_y&1)) * i_stride;
  648.      h->mb.pic.i_stride[i] = i_stride2;
  649.      h->mb.pic.p_fenc_plane[i] = &h->fenc->plane[i][i_pix_offset];
  650. @@ -507,11 +533,28 @@ static void ALWAYS_INLINE x264_macroblock_load_pic_pointers( x264_t *h, int mb_x
  651.          h->mc.load_deinterleave_8x8x2_fenc( h->mb.pic.p_fenc[1], h->mb.pic.p_fenc_plane[1], i_stride2 );
  652.          memcpy( h->mb.pic.p_fdec[1]-FDEC_STRIDE, intra_fdec, 8*sizeof(pixel) );
  653.          memcpy( h->mb.pic.p_fdec[2]-FDEC_STRIDE, intra_fdec+8, 8*sizeof(pixel) );
  654. +        if( h->sh.b_mbaff )
  655. +        {
  656. +            // Top left samples.
  657. +            h->mb.pic.p_fdec[1][-FDEC_STRIDE-1] = h->intra_diagonal_backup[h->mb.topleft_border_index][1][mb_x*9];
  658. +            h->mb.pic.p_fdec[2][-FDEC_STRIDE-1] = h->intra_diagonal_backup[h->mb.topleft_border_index][2][mb_x*9];
  659. +            // Top right samples.
  660. +            CP32( &h->mb.pic.p_fdec[1][-FDEC_STRIDE+8], &h->intra_diagonal_backup[h->mb.topright_border_index][1][mb_x*9+1] );
  661. +            CP32( &h->mb.pic.p_fdec[2][-FDEC_STRIDE+8], &h->intra_diagonal_backup[h->mb.topright_border_index][2][mb_x*9+1] );
  662. +        }
  663. +
  664.      }
  665.      else
  666.      {
  667.          h->mc.copy[PIXEL_16x16]( h->mb.pic.p_fenc[0], FENC_STRIDE, h->mb.pic.p_fenc_plane[0], i_stride2, 16 );
  668. -        memcpy( h->mb.pic.p_fdec[0]-FDEC_STRIDE, intra_fdec, 24*sizeof(pixel) );
  669. +        if( h->sh.b_mbaff )
  670. +        {
  671. +            memcpy( h->mb.pic.p_fdec[0]-FDEC_STRIDE, intra_fdec, 16*sizeof(pixel) );
  672. +            h->mb.pic.p_fdec[0][-FDEC_STRIDE-1] = h->intra_diagonal_backup[h->mb.topleft_border_index][0][mb_x*9];
  673. +            CP64( &h->mb.pic.p_fdec[0][-FDEC_STRIDE+16], &h->intra_diagonal_backup[h->mb.topright_border_index][0][mb_x*9+1] );
  674. +        }
  675. +        else
  676. +            memcpy( h->mb.pic.p_fdec[0]-FDEC_STRIDE, intra_fdec, 24*sizeof(pixel) );
  677.      }
  678.      if( b_interlaced )
  679.      {
  680. @@ -568,6 +611,10 @@ static void inline x264_macroblock_cache_load_neighbours( x264_t *h, int mb_x, i
  681.      h->mb.i_mb_xy = mb_y * h->mb.i_mb_stride + mb_x;
  682.      h->mb.i_b8_xy = 2*(mb_y * h->mb.i_b8_stride + mb_x);
  683.      h->mb.i_b4_xy = 4*(mb_y * h->mb.i_b4_stride + mb_x);
  684. +    h->mb.left_b8[0] =
  685. +    h->mb.left_b8[1] = -1;
  686. +    h->mb.left_b4[0] =
  687. +    h->mb.left_b4[1] = -1;
  688.      h->mb.i_neighbour = 0;
  689.      h->mb.i_neighbour_intra = 0;
  690.      h->mb.i_neighbour_frame = 0;
  691. @@ -580,16 +627,111 @@ static void inline x264_macroblock_cache_load_neighbours( x264_t *h, int mb_x, i
  692.      h->mb.i_mb_type_topleft = -1;
  693.      h->mb.i_mb_type_topright = -1;
  694.      h->mb.left_index_table = left_indices[4];
  695. +    h->mb.topleft_partition = 0;
  696. +    h->mb.topright_border_index =
  697. +    h->mb.topleft_border_index = !(mb_y&1);
  698. +    h->mb.intra_border_index = mb_y&1;
  699. +
  700. +    int topleft = top - 1;
  701. +    int topright = top + 1;
  702. +    int left[2];
  703. +
  704. +    left[0] = left[1] = h->mb.i_mb_xy - 1;
  705. +    h->mb.left_b8[0] = h->mb.left_b8[1] = h->mb.i_b8_xy - 2;
  706. +    h->mb.left_b4[0] = h->mb.left_b4[1] = h->mb.i_b4_xy - 4;
  707. +
  708. +    if( h->sh.b_mbaff )
  709. +    {
  710. +        if( mb_y&1 )
  711. +        {
  712. +            if( mb_x && h->mb.b_interlaced != h->mb.field[h->mb.i_mb_xy-1] )
  713. +            {
  714. +                left[0] = left[1] = h->mb.i_mb_xy - 1 - h->mb.i_mb_stride;
  715. +                h->mb.left_b8[0] = h->mb.left_b8[1] = h->mb.i_b8_xy - 2 - 2*h->mb.i_b8_stride;
  716. +                h->mb.left_b4[0] = h->mb.left_b4[1] = h->mb.i_b4_xy - 4 - 4*h->mb.i_b4_stride;
  717. +
  718. +                if( h->mb.b_interlaced )
  719. +                {
  720. +                    h->mb.left_index_table = left_indices[3];
  721. +                    left[1] += h->mb.i_mb_stride;
  722. +                    h->mb.left_b8[1] += 2*h->mb.i_b8_stride;
  723. +                    h->mb.left_b4[1] += 4*h->mb.i_b4_stride;
  724. +                }
  725. +                else
  726. +                {
  727. +                    h->mb.left_index_table = left_indices[1];
  728. +                    topleft += h->mb.i_mb_stride;
  729. +                    h->mb.topleft_partition = 1;
  730. +                    h->mb.topleft_border_index = 3;
  731. +                }
  732. +            }
  733. +            if( h->mb.b_interlaced )
  734. +                h->mb.topleft_border_index = 1;
  735. +            else
  736. +            {
  737. +                topright = -1;
  738. +                h->mb.intra_border_index = 0;
  739. +            }
  740. +            h->mb.topright_border_index = 1;
  741. +        }
  742. +        else
  743. +        {
  744. +            if( h->mb.b_interlaced )
  745. +            {
  746. +                if( top >= 0 )
  747. +                {
  748. +                    top += h->mb.i_mb_stride*(!h->mb.field[top]);
  749. +                    if( mb_x )
  750. +                        topleft += h->mb.i_mb_stride*(!h->mb.field[topleft]);
  751. +                    if( mb_x < h->mb.i_mb_width )
  752. +                        topright += h->mb.i_mb_stride*(!h->mb.field[topright]);
  753. +                }
  754. +
  755. +                if( top >= 0)
  756. +                {
  757. +                   if( !h->mb.field[top] )
  758. +                    h->mb.intra_border_index = 2;
  759. +
  760. +                    if( topright >=0 && h->mb.field[topright] )
  761. +                        h->mb.topright_border_index = 0;
  762. +                    else
  763. +                        h->mb.topright_border_index = 2;
  764. +
  765. +                    if( topleft >=0 && h->mb.field[topleft] )
  766. +                        h->mb.topleft_border_index = 0;
  767. +                    else
  768. +                        h->mb.topleft_border_index = 2;
  769. +                }
  770. +            }
  771. +            else
  772. +                h->mb.intra_border_index = 1;
  773. +            if( mb_x && h->mb.b_interlaced != h->mb.field[h->mb.i_mb_xy-1] )
  774. +            {
  775. +                if( h->mb.b_interlaced )
  776. +                {
  777. +                    h->mb.left_index_table = left_indices[2];
  778. +                    left[1] += h->mb.i_mb_stride;
  779. +                    h->mb.left_b8[1] += 2*h->mb.i_b8_stride;
  780. +                    h->mb.left_b4[1] += 4*h->mb.i_b4_stride;
  781. +                }
  782. +                else
  783. +                    h->mb.left_index_table = left_indices[0];
  784. +            }
  785. +        }
  786. +    }
  787.  
  788.      if( mb_x > 0 )
  789.      {
  790.          h->mb.i_neighbour_frame |= MB_LEFT;
  791. -        h->mb.i_mb_left_xy[0] = h->mb.i_mb_xy - 1;
  792. +        h->mb.i_mb_left_xy[0] = left[0];
  793. +        h->mb.i_mb_left_xy[1] = left[1];
  794.          h->mb.i_mb_type_left[0] = h->mb.type[h->mb.i_mb_left_xy[0]];
  795. +        h->mb.i_mb_type_left[1] = h->mb.type[h->mb.i_mb_left_xy[1]];
  796.          if( h->mb.i_mb_xy > h->sh.i_first_mb )
  797.          {
  798.              h->mb.i_neighbour |= MB_LEFT;
  799.  
  800. +            // FIXME: We don't currently support constrained intra + mbaff.
  801.              if( !h->param.b_constrained_intra || IS_INTRA( h->mb.i_mb_type_left[0] ) )
  802.                  h->mb.i_neighbour_intra |= MB_LEFT;
  803.          }
  804. @@ -622,12 +764,12 @@ static void inline x264_macroblock_cache_load_neighbours( x264_t *h, int mb_x, i
  805.              }
  806.          }
  807.  
  808. -        if( mb_x > 0 && top - 1 >= 0  )
  809. +        if( mb_x > 0 && topleft >= 0  )
  810.          {
  811.              h->mb.i_neighbour_frame |= MB_TOPLEFT;
  812. -            h->mb.i_mb_topleft_xy = top - 1;
  813. +            h->mb.i_mb_topleft_xy = topleft;
  814.              h->mb.i_mb_type_topleft = h->mb.type[h->mb.i_mb_topleft_xy];
  815. -            if( top - 1 >= h->sh.i_first_mb )
  816. +            if( topleft >= h->sh.i_first_mb )
  817.              {
  818.                  h->mb.i_neighbour |= MB_TOPLEFT;
  819.  
  820. @@ -636,12 +778,12 @@ static void inline x264_macroblock_cache_load_neighbours( x264_t *h, int mb_x, i
  821.              }
  822.          }
  823.  
  824. -        if( mb_x < h->mb.i_mb_width - 1 && top + 1 >= 0 )
  825. +        if( mb_x < h->mb.i_mb_width - 1 && topright >= 0 )
  826.          {
  827.              h->mb.i_neighbour_frame |= MB_TOPRIGHT;
  828. -            h->mb.i_mb_topright_xy = top + 1;
  829. +            h->mb.i_mb_topright_xy = topright;
  830.              h->mb.i_mb_type_topright = h->mb.type[h->mb.i_mb_topright_xy];
  831. -            if( top + 1 >= h->sh.i_first_mb )
  832. +            if( topright >= h->sh.i_first_mb )
  833.              {
  834.                  h->mb.i_neighbour |= MB_TOPRIGHT;
  835.  
  836. @@ -656,9 +798,9 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  837.  {
  838.      x264_macroblock_cache_load_neighbours( h, mb_x, mb_y );
  839.  
  840. -    int left = h->mb.i_mb_left_xy[0];
  841. +    int *left = h->mb.i_mb_left_xy;
  842.      int top  = h->mb.i_mb_top_xy;
  843. -    int top_y = mb_y - (1 << h->mb.b_interlaced);
  844. +    int top_y = top / h->mb.i_mb_stride;
  845.      int s8x8 = h->mb.i_b8_stride;
  846.      int s4x4 = h->mb.i_b4_stride;
  847.      int top_8x8 = (2*top_y+1) * s8x8 + 2*mb_x;
  848. @@ -712,25 +854,25 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  849.  
  850.      if( h->mb.i_neighbour & MB_LEFT )
  851.      {
  852. -        h->mb.cache.i_cbp_left = cbp[left];
  853. +        h->mb.cache.i_cbp_left = cbp[left[0]];
  854.  
  855.          /* load intra4x4 */
  856. -        h->mb.cache.intra4x4_pred_mode[x264_scan8[0 ] - 1] = i4x4[left][left_index_table[0]];
  857. -        h->mb.cache.intra4x4_pred_mode[x264_scan8[2 ] - 1] = i4x4[left][left_index_table[1]];
  858. -        h->mb.cache.intra4x4_pred_mode[x264_scan8[8 ] - 1] = i4x4[left][left_index_table[2]];
  859. -        h->mb.cache.intra4x4_pred_mode[x264_scan8[10] - 1] = i4x4[left][left_index_table[3]];
  860. +        h->mb.cache.intra4x4_pred_mode[x264_scan8[0 ] - 1] = i4x4[left[0]][left_index_table[0]];
  861. +        h->mb.cache.intra4x4_pred_mode[x264_scan8[2 ] - 1] = i4x4[left[0]][left_index_table[1]];
  862. +        h->mb.cache.intra4x4_pred_mode[x264_scan8[8 ] - 1] = i4x4[left[1]][left_index_table[2]];
  863. +        h->mb.cache.intra4x4_pred_mode[x264_scan8[10] - 1] = i4x4[left[1]][left_index_table[3]];
  864.  
  865.          /* load non_zero_count */
  866. -        h->mb.cache.non_zero_count[x264_scan8[0 ] - 1] = nnz[left][left_index_table[4+0]];
  867. -        h->mb.cache.non_zero_count[x264_scan8[2 ] - 1] = nnz[left][left_index_table[4+1]];
  868. -        h->mb.cache.non_zero_count[x264_scan8[8 ] - 1] = nnz[left][left_index_table[4+2]];
  869. -        h->mb.cache.non_zero_count[x264_scan8[10] - 1] = nnz[left][left_index_table[4+3]];
  870. +        h->mb.cache.non_zero_count[x264_scan8[0 ] - 1] = nnz[left[0]][left_index_table[4+0]];
  871. +        h->mb.cache.non_zero_count[x264_scan8[2 ] - 1] = nnz[left[0]][left_index_table[4+1]];
  872. +        h->mb.cache.non_zero_count[x264_scan8[8 ] - 1] = nnz[left[1]][left_index_table[4+2]];
  873. +        h->mb.cache.non_zero_count[x264_scan8[10] - 1] = nnz[left[1]][left_index_table[4+3]];
  874.  
  875. -        h->mb.cache.non_zero_count[x264_scan8[16+0] - 1] = nnz[left][left_index_table[4+4]];
  876. -        h->mb.cache.non_zero_count[x264_scan8[16+2] - 1] = nnz[left][left_index_table[4+5]];
  877. +        h->mb.cache.non_zero_count[x264_scan8[16+0] - 1] = nnz[left[0]][left_index_table[4+4]];
  878. +        h->mb.cache.non_zero_count[x264_scan8[16+2] - 1] = nnz[left[1]][left_index_table[4+5]];
  879.  
  880. -        h->mb.cache.non_zero_count[x264_scan8[16+4+0] - 1] = nnz[left][left_index_table[4+6]];
  881. -        h->mb.cache.non_zero_count[x264_scan8[16+4+2] - 1] = nnz[left][left_index_table[4+7]];
  882. +        h->mb.cache.non_zero_count[x264_scan8[16+4+0] - 1] = nnz[left[0]][left_index_table[4+6]];
  883. +        h->mb.cache.non_zero_count[x264_scan8[16+4+2] - 1] = nnz[left[1]][left_index_table[4+7]];
  884.      }
  885.      else
  886.      {
  887. @@ -755,7 +897,7 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  888.      if( h->pps->b_transform_8x8_mode )
  889.      {
  890.          h->mb.cache.i_neighbour_transform_size =
  891. -            ( (h->mb.i_neighbour & MB_LEFT) && h->mb.mb_transform_size[left] )
  892. +            ( (h->mb.i_neighbour & MB_LEFT) && h->mb.mb_transform_size[left[0]] )
  893.            + ( (h->mb.i_neighbour & MB_TOP) && h->mb.mb_transform_size[top]  );
  894.      }
  895.  
  896. @@ -768,7 +910,7 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  897.            + !!(h->mb.i_neighbour & MB_TOP);
  898.      }
  899.  
  900. -    if( !h->mb.b_interlaced )
  901. +    if( !h->sh.b_mbaff )
  902.      {
  903.          x264_copy_column8( h->mb.pic.p_fdec[0]-1+ 4*FDEC_STRIDE, h->mb.pic.p_fdec[0]+15+ 4*FDEC_STRIDE );
  904.          x264_copy_column8( h->mb.pic.p_fdec[0]-1+12*FDEC_STRIDE, h->mb.pic.p_fdec[0]+15+12*FDEC_STRIDE );
  905. @@ -869,10 +1011,10 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  906.  
  907.              if( h->mb.i_neighbour & MB_LEFT )
  908.              {
  909. -                CP16( h->mb.cache.mvd[l][x264_scan8[0 ] - 1], mvd[left][left_index_table[0]] );
  910. -                CP16( h->mb.cache.mvd[l][x264_scan8[2 ] - 1], mvd[left][left_index_table[1]] );
  911. -                CP16( h->mb.cache.mvd[l][x264_scan8[8 ] - 1], mvd[left][left_index_table[2]] );
  912. -                CP16( h->mb.cache.mvd[l][x264_scan8[10] - 1], mvd[left][left_index_table[3]] );
  913. +                CP16( h->mb.cache.mvd[l][x264_scan8[0 ] - 1], mvd[left[0]][left_index_table[0]] );
  914. +                CP16( h->mb.cache.mvd[l][x264_scan8[2 ] - 1], mvd[left[0]][left_index_table[1]] );
  915. +                CP16( h->mb.cache.mvd[l][x264_scan8[8 ] - 1], mvd[left[1]][left_index_table[2]] );
  916. +                CP16( h->mb.cache.mvd[l][x264_scan8[10] - 1], mvd[left[1]][left_index_table[3]] );
  917.              }
  918.              else
  919.                  for( int i = 0; i < 4; i++ )
  920. @@ -889,7 +1031,7 @@ void x264_macroblock_cache_load( x264_t *h, int mb_x, int mb_y )
  921.          {
  922.              uint8_t skipbp;
  923.              x264_macroblock_cache_skip( h, 0, 0, 4, 4, 0 );
  924. -            skipbp = (h->mb.i_neighbour & MB_LEFT) ? h->mb.skipbp[left] : 0;
  925. +            skipbp = (h->mb.i_neighbour & MB_LEFT) ? h->mb.skipbp[left[0]] : 0;
  926.              h->mb.cache.skip[x264_scan8[0] - 1] = skipbp & 0x2;
  927.              h->mb.cache.skip[x264_scan8[8] - 1] = skipbp & 0x8;
  928.              skipbp = (h->mb.i_neighbour & MB_TOP) ? h->mb.skipbp[top] : 0;
  929. @@ -968,11 +1110,11 @@ void x264_macroblock_cache_load_deblock( x264_t *h )
  930.  
  931.              if( h->mb.i_neighbour & MB_LEFT )
  932.              {
  933. -                int left = h->mb.i_mb_left_xy[0];
  934. -                h->mb.cache.non_zero_count[x264_scan8[0 ] - 1] = nnz[left][left_index_table[4+0]];
  935. -                h->mb.cache.non_zero_count[x264_scan8[2 ] - 1] = nnz[left][left_index_table[4+1]];
  936. -                h->mb.cache.non_zero_count[x264_scan8[8 ] - 1] = nnz[left][left_index_table[4+2]];
  937. -                h->mb.cache.non_zero_count[x264_scan8[10] - 1] = nnz[left][left_index_table[4+3]];
  938. +                int *left = h->mb.i_mb_left_xy;
  939. +                h->mb.cache.non_zero_count[x264_scan8[0 ] - 1] = nnz[left[0]][left_index_table[4+0]];
  940. +                h->mb.cache.non_zero_count[x264_scan8[2 ] - 1] = nnz[left[0]][left_index_table[4+1]];
  941. +                h->mb.cache.non_zero_count[x264_scan8[8 ] - 1] = nnz[left[1]][left_index_table[4+2]];
  942. +                h->mb.cache.non_zero_count[x264_scan8[10] - 1] = nnz[left[1]][left_index_table[4+3]];
  943.              }
  944.  
  945.              for( int l = 0; l <= (h->sh.i_type == SLICE_TYPE_B); l++ )
  946. @@ -1103,20 +1245,80 @@ static void ALWAYS_INLINE x264_macroblock_store_pic( x264_t *h, int mb_x, int mb
  947.      int i_pix_offset = b_interlaced
  948.                       ? 16 * mb_x + w * (mb_y&~1) * i_stride + (mb_y&1) * i_stride
  949.                       : 16 * mb_x + w * mb_y * i_stride;
  950. +    const int intra_diag_width = 8+1; // One top left sample, then eight top right samples.
  951.      pixel *intra_fdec = &h->intra_border_backup[mb_y&1][i][mb_x*16];
  952.      if( i )
  953.      {
  954. +        if( h->sh.b_mbaff )
  955. +        {
  956. +            /* Frame macroblocks use the macroblock directly above for intra
  957. +             * prediction. Field macroblock pairs predict from fields of the same
  958. +             * parity. However field macroblock pairs predicting from frame pairs
  959. +             * use the bottom two rows of the frame for prediction, the penultimate
  960. +             * row is stored in intra_border_backup[2]. */
  961. +            if( mb_y&1 )
  962. +            {
  963. +                if( mb_x )
  964. +                {
  965. +                    // Store top left.
  966. +                    h->intra_diagonal_backup[1][1][mb_x*intra_diag_width] = h->intra_border_backup[1][1][(mb_x-1)*16+7];
  967. +                    h->intra_diagonal_backup[1][2][mb_x*intra_diag_width] = h->intra_border_backup[1][1][(mb_x-1)*16+8+7];
  968. +                    h->intra_diagonal_backup[2][1][mb_x*intra_diag_width] = h->intra_border_backup[2][1][(mb_x-1)*16+7];
  969. +                    h->intra_diagonal_backup[2][2][mb_x*intra_diag_width] = h->intra_border_backup[2][1][(mb_x-1)*16+8+7];
  970. +                    // Store top right.
  971. +                    CP32( &h->intra_diagonal_backup[1][1][(mb_x-1)*intra_diag_width+1], h->mb.pic.p_fdec[0]+FDEC_STRIDE*7 );
  972. +                    CP32( &h->intra_diagonal_backup[1][2][(mb_x-1)*intra_diag_width+1], h->mb.pic.p_fdec[0]+FDEC_STRIDE*7 );
  973. +                    CP32( &h->intra_diagonal_backup[2][1][(mb_x-1)*intra_diag_width+1], h->mb.pic.p_fdec[0]+FDEC_STRIDE*6 );
  974. +                    CP32( &h->intra_diagonal_backup[2][2][(mb_x-1)*intra_diag_width+1], h->mb.pic.p_fdec[0]+FDEC_STRIDE*6 );
  975. +                }
  976. +                memcpy( &h->intra_border_backup[2][i][mb_x*16],   h->mb.pic.p_fdec[1]+FDEC_STRIDE*6, 8*sizeof(pixel) );
  977. +                memcpy( &h->intra_border_backup[2][i][mb_x*16]+8, h->mb.pic.p_fdec[2]+FDEC_STRIDE*6, 8*sizeof(pixel) );
  978. +            }
  979. +            else
  980. +            {
  981. +                if( mb_x )
  982. +                {
  983. +                    h->intra_diagonal_backup[0][1][mb_x*intra_diag_width] = h->intra_border_backup[0][1][(mb_x-1)*16+7];
  984. +                    h->intra_diagonal_backup[0][2][mb_x*intra_diag_width] = h->intra_border_backup[0][1][(mb_x-1)*16+8+7];
  985. +                    CP32( &h->intra_diagonal_backup[0][1][(mb_x-1)*intra_diag_width+1], h->mb.pic.p_fdec[1]+FDEC_STRIDE*7 );
  986. +                    CP32( &h->intra_diagonal_backup[0][2][(mb_x-1)*intra_diag_width+1], h->mb.pic.p_fdec[2]+FDEC_STRIDE*7 );
  987. +                }
  988. +                // Sometimes needed for bottom macroblock of this pair.
  989. +                h->intra_diagonal_backup[3][1][mb_x*intra_diag_width] = h->mb.pic.p_fdec[1][-1+7*FDEC_STRIDE];
  990. +                h->intra_diagonal_backup[3][2][mb_x*intra_diag_width] = h->mb.pic.p_fdec[2][-1+7*FDEC_STRIDE];
  991. +            }
  992. +        }
  993.          h->mc.store_interleave_8x8x2( &h->fdec->plane[1][i_pix_offset], i_stride2, h->mb.pic.p_fdec[1], h->mb.pic.p_fdec[2] );
  994.          memcpy( intra_fdec,   h->mb.pic.p_fdec[1]+FDEC_STRIDE*7, 8*sizeof(pixel) );
  995.          memcpy( intra_fdec+8, h->mb.pic.p_fdec[2]+FDEC_STRIDE*7, 8*sizeof(pixel) );
  996. -        twiddle_topleft_pixel( h->mb.pic.p_fdec[1]-FDEC_STRIDE-1, h->mb.pic.p_fdec[1]-FDEC_STRIDE+7, b_interlaced );
  997. -        twiddle_topleft_pixel( h->mb.pic.p_fdec[2]-FDEC_STRIDE-1, h->mb.pic.p_fdec[2]-FDEC_STRIDE+7, b_interlaced );
  998. +        twiddle_topleft_pixel( h->mb.pic.p_fdec[1]-FDEC_STRIDE-1, h->mb.pic.p_fdec[1]-FDEC_STRIDE+7, h->sh.b_mbaff );
  999. +        twiddle_topleft_pixel( h->mb.pic.p_fdec[2]-FDEC_STRIDE-1, h->mb.pic.p_fdec[2]-FDEC_STRIDE+7, h->sh.b_mbaff );
  1000.      }
  1001.      else
  1002.      {
  1003. +        if( h->sh.b_mbaff )
  1004. +        {
  1005. +            if( mb_y&1 )
  1006. +            {
  1007. +                if( mb_x )
  1008. +                {
  1009. +                    h->intra_diagonal_backup[1][0][mb_x*intra_diag_width] = h->intra_border_backup[1][0][(mb_x-1)*16+15];
  1010. +                    h->intra_diagonal_backup[2][0][mb_x*intra_diag_width] = h->intra_border_backup[2][0][(mb_x-1)*16+15];
  1011. +                    CP64( &h->intra_diagonal_backup[1][0][(mb_x-1)*intra_diag_width+1], h->mb.pic.p_fdec[0]+FDEC_STRIDE*15 );
  1012. +                    CP64( &h->intra_diagonal_backup[2][0][(mb_x-1)*intra_diag_width+1], h->mb.pic.p_fdec[0]+FDEC_STRIDE*14 );
  1013. +                }
  1014. +                memcpy( &h->intra_border_backup[2][i][mb_x*16], h->mb.pic.p_fdec[0]+FDEC_STRIDE*14, 16*sizeof(pixel) );
  1015. +            }
  1016. +            else
  1017. +            {
  1018. +                h->intra_diagonal_backup[0][0][mb_x*intra_diag_width] = h->intra_border_backup[0][0][(mb_x-1)*16+15];
  1019. +                CP64( &h->intra_diagonal_backup[0][0][(mb_x-1)*intra_diag_width+1], h->mb.pic.p_fdec[0]+FDEC_STRIDE*15 );
  1020. +                h->intra_diagonal_backup[3][0][mb_x*intra_diag_width] = h->mb.pic.p_fdec[0][-1+15*FDEC_STRIDE];
  1021. +            }
  1022. +        }
  1023.          h->mc.copy[PIXEL_16x16]( &h->fdec->plane[0][i_pix_offset], i_stride2, h->mb.pic.p_fdec[0], FDEC_STRIDE, 16 );
  1024.          memcpy( intra_fdec, h->mb.pic.p_fdec[0]+FDEC_STRIDE*15, 16*sizeof(pixel) );
  1025. -        twiddle_topleft_pixel( h->mb.pic.p_fdec[0]-FDEC_STRIDE-1, h->mb.pic.p_fdec[0]-FDEC_STRIDE+15, b_interlaced );
  1026. +        twiddle_topleft_pixel( h->mb.pic.p_fdec[0]-FDEC_STRIDE-1, h->mb.pic.p_fdec[0]-FDEC_STRIDE+15, h->sh.b_mbaff );
  1027.      }
  1028.  }
  1029.  
  1030. --
  1031. 1.7.1
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!